A Study On Spatial Similarity Theory And Calculation Model

Spatial Similarity is an important theoretic problem in GIS. There are few researches on spatial similarity at home and abroad. The main reasons lie in two aspects: One reason is that it is not convenient to calculate the similarity. The other reason is that the similarity could show the deeper information than simple spatial data. It is very difficult to calculate spatial similarity. Spatial problems can be solved using similar spatial phenomena. The similarity can be utilized to classify objects, to form concepts, to solve problems, and to make generalizations. Spatial similarity would help explain certain phenomena and their surrounding circumstances. Which help of the method of ontology and cognition, the similarity for spatial directions, the similarity for spatial topological relations, the semantic similarity, the similarity for spatial scenes are researched, and their calculation methods are given.Spatial relations could be described using some data models, could be used to represent mutual relationships between spatial objects with a certain location, attribute and configuration. It is a kind of relation between spatial objects which forms in some regions, and relates to spatial features. Spatial relations are divided into spatial direction relation, spatial topological relation and spatial distance relation. Spatial direction relation is described by quantitative and qualitative methods. In a quantitative description, the azimuth is used to represent direction relation. And in a qualitative description, the ordinal data is used to represent direction relation. Moving distances of the comparative object to a reference object could describe the direction relation between two objects. If the comparative object moves to a different direction to the reference object, the spatial direction relation between the two objects has changed. Whereas, there is similarity between the two direction relations, we should keep the direction relation to similar in Map generalization.The main problem in research on spatial data model is how to preserve topological information between spatial objects as possible when describing the configuration of spatial objects. The change of spatial metric and the reduction of size are related to the simplification of shape of objects, the change of spatial topological relations is related to some objects, or the deletion and amalgamation operation, and dimensional change of parts of these objects. Many map generalization algorithm pay attention to geometric simplification, but ignore the topological attributes and topological relations of these objects, which should be preserved as possible in the process of simplification. This refers to the problems of spatial similarity between topological relations, which is the contents in this dissertation.In GIS, the assessment of semantic similarity is very important. The spatial data are so many that the retrievement and generalization of spatial information become the basic components of GIS at present. We should use better information retrievement methods and generalization mechanism to enhance the usability,which could make users gain expected information. Semantic similarity is mainly used to spatial abstraction and amalgamation of spatial objects. Researches on it are helpful to classify spatial phenomena. We could enhance the usability of geographical information, use familiar geographical space to represent unfamiliar geographical space through semantic similarity. For all above, semantic similarity should be studied deeply.There are many spatial objects in spatial region. All of them would affect the people's spatial cognition. In map generalization, they would be transformed into a different map with different generalization methods. Which objects should be kept or deleted? An important factor is that the generalized map should be similar to original map. How to judge the similarity between two maps? We should investigate the similarity between spatial scenes.After the fields of similarity are analyzed, the spatial similarity theories are described, the calculation models are researched, and then the new methods which calculate spatial similarity are presented in the dissertation. The main research contents and new calculation methods in the dissertation are follows:1. The spatial relations are defined and classified, the theories of spatial similarity are described, and interrelated application fields are summarized.2. Introducing the ontology theory and cognitive rule as a tool for spatial relation classification. Ontology could be used to solve the geographical classification and mutual relationship. Ontology plays an important role in GIS and is the basis of the research on semantic similarity. The cognitive rule is another basis of the spatial similarity theory. Mental map is also called cognitive map. Cognitive map is similar to reality world. Thus it can be seen; the research on spatial similarity has significance.3. Several methods describing spatial direction relations are analyzed. Goyal derived the spatial direction similarity considering the whole object. The method is fit for the processing of vector data. Using the direction-relation matrix model to describe the spatial directions, using the transformation of raster cell direction distance of objects to computer direction similarity, the author gives two new methods of calculating similarity for spatial directions between areal objects in raster data: One is based on the features of raster data and the change of direction distance between areal objects to calculate the similarity for spatial directions, the other is based on the variation of each raster cell angle to calculate the similarity between spatial directions. Because the similarity definition and criterion of the two methods is different, the similarity value is different. The two methods overcome the complexity of calculating similarity for spatial directions based on direction matrix model, and solve the limitation of small change of direction. The two methods can be learned from others' strong points to offset one's weakness, and have broader applicability. The scale is an important feature of spatial data. The changing of the scale would cause the changing of direction between the objects. We should research the similarity between the spatial directionsunder the different scales.4. The similarity between spatial topological relations is summarized. Based on the combinational representation of spatial topological relations, how to combine basic spatial topological relations is analyzed in this thesis. According to the combinational method, two kinds of systems of spatial topological relations between two areas are given, in addition, a reasoning table about spatial topological relations is given, and a complete diagram about spatial topological relations between two areas is listed. The author advanced 21 kinds of topological relations between two areal objects, and 54 kinds of global spatial topological relations between two areas based on basic spatial relation. According to the 21 kinds of topological relations graphs, the author draws concept neighborhood graphs to calculate the topology similarity. The author researched the similarity between different classes of topology, and also researches the similarity between different sub- classes in same class topological relation. The topological relation similarity values by means of 0 dimension intersection and 1 dimension intersection were calculated.5. There has been an increasing interest in semantic similarity in the field of GIS abroad. Semantic similarity application in GIS is to contract and incorporate objects. It is an important part of spatial similarity. In the dissertation, the general concept of semantic similarity is outlined and the traditional measurement methods of semantic similarity are researched. Based on Nominal, Ordinal, Interval and Ratio data, the author advances the nominal scale semantic similarity and non-nominal scale semantic similarity, and proposes a new method to measure the semantic similarity for the features of spatial objects.6. Except for direction relation, topological relation and distance relation transformation factors, there are many factors which affect the spatial similarity together. The similarity between scenes concentrates on the integration of directional, topological, semantic, distance, and the number of the scenes factors.In this paper, three aspects about spatial object similarity are studied, i.e., spatial direction similarity, spatial topological similarity and semantic similarity. In order to investigate spatial scene similarity, the three aspects are combined with similarity of spatial distance, number of level of spatial objects and number of sub-objects of each level. After referring many technical literature and deeply analyzing development of this research presently at home and abroad, the author investigated existing methods and presented new calculative methods, and used ontology and spatial cognitive approach to find out essential characteristic of spatial similarity in the research.This paper aims at exploring theories of spatial similarity and constructing some calculative model, and validating the feasibility of realizing technique and methods. Author expects that more researchers would pay attentions to the field, and more outstanding achievements would be made out in the future.